AI Is Learning to Write Genetic Code

This sort of research is both exciting and terrifying:

The two models in question were told to generate complete genomes for a viable bacteriophage—a type of virus able to infect and replicate itself inside bacteria, destroying them from the inside.

Using an existing bacteriophage as an example—ΦX174 (pronounced “fie-ex-1-7-4”), known for its ability to infect and destroy E. coli bacteria—the models generated about 700,000 potential designs, of which the researchers picked 285 that looked most promising.

The researchers then synthesised new DNA molecules using those designs and inserted them into E. coli bacteria, before waiting to see if viable bacteriophages would emerge.

Shortly afterwards, 16 of the Petri dishes in which the bacteria were growing began to show clear spots, as the viruses began to attack and replicate themselves inside the E. coli, demonstrating their viability.

Some of those viable viruses proved more effective at attacking E. coli than the original ΦX174 bacteriophage.

That’s a positive use of a synthetic virus. We can all imagine the negative uses.

Posted on August 21, 2026 at 12:51 PM10 Comments

Comments

Kempton August 21, 2026 2:03 PM

Hmmm, me too, //both exciting and terrifying// +1
So I’ve been following research on bacteriophage for some years even I only reached out to ask for help days before my dad’s passing. So I’ve reached out to a friend who is a leading researcher re bacteriophage and see what she thinks about this development. Will see what she says. I wish I’m only excited but I won’t be telling the truth if I don’t say I’m 40%+ terrified.

Anonymous August 21, 2026 6:13 PM

Hmmm…
16 out of 700’000 is 0.00228%. I wonder if random mutations on the same DNA would provide similar results?

Paul August 22, 2026 3:12 AM

It might become a great tool in the fight against infectious disease.
At the moment the future of antibiotics is bleak, bacteria rapidly develop resistance to antibiotics.

Rontea August 22, 2026 9:05 AM

To forge a virus that never knew nature is to write a poem in fire and expect the flame to bow.

Andrew Burday August 24, 2026 9:11 PM

The up- and downsides of genetic engineering have been well understood for around 40 years now. Is there any reason to think that AI will make it easier to design either beneficial or harmful organisms? The linked article doesn’t give any. Narrowing the field from 700,000 candidates to 285 must have taken significant human effort, and then only 16 of 285 actually worked. I have no idea what work would have had to be done to produce those 16 without AI, but it’s not obviously more than what these researchers had to put in. (And producing all 16 is a high bar; all you need is one.)

Writing pro-AI propaganda rather than serious analysis isn’t only a matter of making AI sound more beneficial than it actually is. It can also be a matter of making AI sound more important than it actually is. One of the goals of the big AI companies is to overawe the public, so as to grab legal and political benefits such as the ability to violate others’ copyrights and avoid responsibility for their own actions. (E.g. “Mythos used social engineering”, not “Anthropic used social engineering”.) That is propaganda, not analysis.

lurker August 24, 2026 11:28 PM

@Andrew Burday

(E.g. “Mythos used social engineering”, not “Anthropic used social engineering”.) That is propaganda, not analysis.

+1
AFAICT the AI hasn’t taken over yet, so it should be humans in the driver’s seat, and humans who take the rap.

Leave a comment

Blog moderation policy

Login

Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via https://michelf.ca/projects/php-markdown/extra/

Sidebar photo of Bruce Schneier by Joe MacInnis.